Can I use this to validate JSON output from a model?

Absolutely. It includes capabilities for structured output validation, helping you assert that LLM responses comply with your specific JSON schemas.

Does this skill support integration testing for AI?

Yes, it includes specific patterns for using VCR.py to record and replay HTTP interactions, ensuring your integration tests remain deterministic and fast.

How does this prevent flaky tests in AI development?

It enforces strict rules against live API calls and random seeds, replacing them with mocking strategies and controlled fixtures to ensure repeatable results.

How are async LLM calls handled in these patterns?

The skill provides patterns for async timeout testing and polling assertions, ensuring your application handles slow or timed-out AI requests gracefully.

Which LLM evaluation frameworks are supported?

The skill provides built-in support and code patterns for DeepEval and RAGAS, allowing you to measure metrics like faithfulness, relevancy, and hallucination rates.

LLM Application Testing

Name: LLM Application Testing
Author: yonatangross

byyonatangross

•

보안 및 테스팅

Implements deterministic testing patterns for AI applications using DeepEval, RAGAS, and advanced mocking strategies.

This skill provides a comprehensive framework for testing LLM-based applications within Claude Code, ensuring reliability and performance through standardized, deterministic patterns. It streamlines the creation of unit and integration tests by offering production-ready approaches for mocking LLM responses, validating structured JSON outputs, and implementing sophisticated quality metrics like faithfulness and relevancy via DeepEval and RAGAS. By enforcing strict anti-patterns—such as avoiding live API calls in CI and requiring explicit timeouts—it empowers developers to build robust AI systems with reliable regression testing, golden dataset comparisons, and high-confidence deployment cycles.

주요 기능

01Integrate DeepEval and RAGAS for automated quality metric assertions

02Validate structured outputs and JSON schema compliance for AI responses

0369 GitHub stars

04Mock LLM responses for deterministic, cost-effective unit testing

05Implement VCR.py recording for reliable integration test suites

06Enforce async timeout handling and edge case validation for AI workflows

사용 사례

01Running CI/CD pipelines for AI apps without incurring live API costs

02Evaluating RAG system performance using relevancy and faithfulness metrics

03Testing application resilience against high-latency LLM responses and timeouts

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yonatangross/orchestkit llm-testing

For use in Claude.ai and ChatGPT

Download Skill